fix rmw_get_graph_guard_condition

[wjwwood]

Previously it was just returning junk, which was left over from my wait_for_service work.

This doesn't implement it, but at least it returns a valid guard condition. I wouldn't return anything at all, but rcl_node_init calls it, so I have to return something valid or else all fastrtps programs would fail.

[wjwwood]

CI:

• Linux:
  • 
• OS X:
  • 
• Windows:
  • 

[tfoote]

lgtm, there are 6 instances of the 7 lines to get the participant from the impl that could be turned into a function. If there's expected to be more instances the helper function might be worth it. Otherwise we can leave it as is.

[wjwwood]

Yeah, there's a lot of improvements that could be made, but with respect to deduplicating the code, in about half the cases you have to return NULL and the other half return RMW_RET_ERROR if there is a problem. This is why we have common macros to encapsulate this in rmw:

https://github.com/ros2/rmw/blob/master/rmw/include/rmw/impl/cpp/macros.hpp

And in rcl as well:

https://github.com/ros2/rcl/blob/master/rcl/src/rcl/common.h#L26-L33

I'd leave it as-is for now and try to deduplicate code between this and the other rmw implementations in the future.

[wjwwood]

CI looks good (existing warnings on Windows and flaky tests on Linux).

[dirk-thomas]

Doesn't this leak graph_guard_condition?

Same for the other return statements below.

[wjwwood]

Yes. The function was not setup to handle unrolling. None of the functions in this file are except one. So I didn't want to change that in these changes. In general I think this time implementation needs some clean up.

[wjwwood]

#33